Grid Energy Logic
Short-horizon nodal signals and grid analytics—grounded in wholesale pricing (PJM) and meteorological context (NOAA)—exposed through auditable API contracts.
Problem framing
Wholesale electricity is not a single scalar: it is a high-dimensional field over pricing nodes, shaped by unit commitment, reserve products, transmission limits, and loss models. At sub-hourly resolution, LMP trajectories exhibit jumps, mean reversion, and cross-node dependence that naive spot-temperature regressions fail to capture.
Meanwhile, radiation and cloud fields modulate distributed generation and cooling load, linking mesoscale meteorology to balancing stress within minutes to hours. Any credible short-horizon oracle must therefore fuse ISO price discovery with numerically stable weather features—not folklore correlations.
Sketch of the conditioning structure (informal)
Notation is illustrative. Estimators, regularity conditions, and scoring rules belong in your model card and backtest appendix—not in this sketch.
GRENL separates concerns the way serious infrastructure teams do: a focused API service for data contracts and inference, typed Python boundaries, and automated quality gates. The landing surface you are reading is intentionally thin—serious logic ships in core-api.
Map human-readable hubs and zones to PJM pricing nodes; serve consistent identifiers so estimators and risk systems agree on the same pnode key.
Hot-path LMP materialized in Redis with explicit TTL semantics—so bursty agent fleets do not stampede primary ISO feeds.
Optional x402-style payment headers for machine-to-machine metering; current public implementation is a placeholder gate designed for integration testing.
Interactive
Tune toy parameters, drag levers on a synthetic trace, and skim guided cards—nothing here calls production services; it is all in-browser for intuition.
Synthetic signal
Primary curve is a synthetic imbalance proxy; cyan curve is a transformed response after weather and horizon conditioning. It is structured, but not a validated market forecast.
Imbalance
58
Ramp risk
62
Spread
±35 MW
Play with scalars
Drag the controls—notation updates in real time. Values are illustrative only; they do not connect to production models.
Implied output
40.3 a.u.
Sketch confidence
65%
Q-hat(tau, p, omega, theta) is estimated from price state, weather coupling, and calibration settings.
tau=30 min, theta=0.42, omega=0.65
Boxed values follow the sliders; the top line is fixed structure.
Guided tour
Use arrows, dots, or keyboard ← →. Each card is a compressed narrative—deep dives live in integration docs and onboarding materials.
Markets
Realtime nodal prices fold in congestion, losses, and commitment. GRENL treats ISO feeds as the primary discovery channel and maps human-readable hubs to stable pnode keys so agents share one coordinate system.
Technical depth
Select a focus area. Copy is written for readers comfortable with ISO markets, numerical weather prediction, and basic statistical decision theory.
Organized wholesale markets clear at locational marginal prices (LMPs)—nodal valuations that aggregate generator offers, load bids, and the shadow prices of binding transmission constraints. In PJM-class footprints, realtime LMP therefore encodes congestion rents, loss allocation, and commitment outcomes at the pricing node.
GRENL's ingestion path treats ISO-published realtime quotes as the primary price-discovery channel for short-horizon inference, with explicit mapping from human-readable hubs to pnode identifiers so downstream estimators see a stable coordinate system across sessions.
Lens selector
Choose how you want the narrative framed. This does not change product behavior—it reorganizes what we highlight for your role.
Emphasis on nodal LMP as a high-frequency state variable, cross-node coherence, and deterministic signals suitable for execution logic and risk limits—not hand-wavy “AI alpha.”
Provenance
We bias toward sources you can cite in a methods section: ISO primary feeds, NOAA open data, and durable relational stores—not screenshots or unversioned spreadsheets.
Realtime and archival LMP series as published by the ISO. Authentication via subscription key; rate limits treated as part of the control problem, not an afterthought.
Mesoscale analyses and short-range forecasts from NOAA’s open dissemination pathways (e.g., Registry of Open Data on AWS). GRIB/Zarr pipelines favor reproducible, citable inputs over opaque scrapers.
Runtime quote cache uses explicit TTL (300s) in Redis. The core API data layer is Postgres/Timescale-compatible for deployment-specific historical retention policies.
Data governance
Public behavior is explicit: every source has a cadence, a retention window, and deterministic error surfacing. No silent fallbacks.
Security posture
Default API mode is contract-driven access with optional x402-style payment gate when PAYMENTS_ENABLED is enabled for a deployment.
Provider credentials (for example PJM subscription keys) are server-side environment secrets. Keys are never shipped in browser bundles.
Log verbosity is environment-configurable. Operational logs focus on request outcomes and errors, with a policy to avoid secret material in logs.
No blanket SOC 2 / ISO 27001 attestation is claimed on this page. Control requirements are handled in customer-specific onboarding and agreements.
Formal trust targets and support response windows are summarized on the Trust page.
Async-first FastAPI workers, structured logging hooks, and container builds verified in CI. The stack is chosen for latency, auditability, and operational simplicity—not novelty for its own sake.
FAQ
Integration
Tell us your deployment profile and we will return an onboarding plan with API scope, expected limits, and rollout checkpoints.